The overall goal of this project continues to be the development and investigation of methods for the efficient design and analysis of clinical studies. A major specific aim will be on the development of methods appropriate for phase III cancer clinical trials, especially multi-arm phase III trials. We will continue to work on statistics appropriate for ordered alternatives, and will develop designs appropriate for situations when the alternatives can be assumed to have such an order restriction. Practical approaches for factorial designs, making explicit allowance for the possibility of interactions, will be developed. We will also continue our work in exploratory methods for survival analysis, emphasizing smoothed regression functions and extensions of the regression tree methodology. We will also continue to work on methods for use with multistate and multivariate survival data, starting with multistate situations which can be summarized by the probability of being in a given state over time (possibly conditional on being alive), and with bivariate survival data subject to univariate censoring. Extensions will be explored as the project proceeds. Other topics that arise in our work on clinical studies will also be explored. Group sequential ideas will be extended to the problem of developing a batted of prognostic factors over time, and small sample distributions for the logrank test will be investigated. Taken as a whole, the project will contribute to improvements in cancer mortality through better methods for design, conduct and analysis of clinical studies.